CN111984355A

CN111984355A - Method and device for realizing man-machine multi-turn conversation

Info

Publication number: CN111984355A
Application number: CN202010844657.1A
Authority: CN
Inventors: 陶阔; 韩燕�
Original assignee: 4Paradigm Beijing Technology Co Ltd
Current assignee: 4Paradigm Beijing Technology Co Ltd
Priority date: 2020-08-20
Filing date: 2020-08-20
Publication date: 2020-11-24
Also published as: WO2022037019A1

Abstract

A method and a device for realizing multiple rounds of human-computer conversations are provided. The method comprises the following steps: providing a user configuration interface, wherein a plurality of types of intention nodes are provided in the user configuration interface; constructing an intention unit based on operations of a first user for selecting an intention node, configuring the intention node and connecting the intention node on the user configuration interface; and realizing man-machine multi-turn conversation with the second user based on the intention unit so as to obtain the key intention of the second user in the man-machine multi-turn conversation and realize judgment and prediction of the conversation.

Description

Method and device for realizing man-machine multi-turn conversation

Technical Field

The invention relates to the field of artificial intelligence, in particular to a method and a device for realizing man-machine multi-turn conversation.

Background

With the development of artificial intelligence technology, the application of man-machine multi-turn conversations is also more and more extensive, such as intelligent conversation robots for reserving hotels, air tickets, and the like. The man-machine multi-turn dialog aims at acquiring necessary information through a series of pursuits after acquiring a preliminary intention of a user to finally provide information and services corresponding to the intention of the user.

In a multi-turn conversation, a user usually has a preliminary intention and wants to obtain information or services (e.g., hotel booking, ticket booking, air order, information consultation, etc.) corresponding to his or her own intention. However, in the face of a complex conversation situation of a user, an intelligent conversation robot program is often difficult to flexibly deal with, cannot effectively acquire key intentions of the user, and cannot provide accurate information and services.

Disclosure of Invention

The invention aims to provide a method and a device for realizing man-machine multi-turn conversation.

One aspect of the present invention provides a method for implementing multiple rounds of human-computer conversations, the method comprising: providing a user configuration interface, wherein a plurality of types of intention nodes are provided in the user configuration interface; constructing an intention unit based on operations of a first user for selecting an intention node, configuring the intention node and connecting the intention node on the user configuration interface; and realizing man-machine multi-turn conversation with the second user based on the intention unit so as to obtain the key intention of the second user in the man-machine multi-turn conversation and realize judgment and prediction of the conversation.

Optionally, the plurality of types of intent nodes include one or more of: the system comprises an input node, a judgment node, an output node, an assignment node and an interface node.

Optionally, the configuration items of the input node include: a trigger question for triggering the task and a question similar to the trigger question; a slot position, wherein the slot position is set based on a trigger question method and a question method similar to the trigger question method; and the task parameters are set based on the triggering question and/or the similar question to the triggering question.

Optionally, the configuration item of the input node further comprises a jump question method for jumping to other input nodes.

Optionally, the configuration item of the judgment node includes parameter logic, where the parameter logic performs logical condition judgment on the value obtained by the previous node and a known value or a custom value, and determines a node next to the judgment node based on the mutually exclusive judgment result.

Optionally, the configuration item of the parameter logic includes a custom judgment value, a custom logic relationship and a custom reference value, wherein the parameter logic is configured to determine whether the custom judgment value and the custom reference value satisfy the custom logic relationship, and determine a next node of the judgment node based on a result of the determination.

Optionally, the customized judgment value includes at least one of a task parameter value in the data received from the previous node, a parameter value in a parameter dictionary, and an interface return value, wherein the customized reference value includes at least one of a set value, an interface return value, and a customized value in the parameter dictionary, and wherein the customized logical relationship is a logical relationship between the customized judgment value selected by the first user according to the intention node and the customized reference value.

Optionally, the configuration item of the output node includes a configuration item that outputs the reply content to the second user in a predetermined form.

Optionally, the textbox in the default state of the content editor of the output node supports adding the obtained task parameters and the output parameters of the interface node.

Optionally, the configuration item of the assignment node includes at least one of a parameter reset and a custom assignment, wherein the parameter reset includes an alternation of slot information based on data received from a node immediately preceding the assignment node, and the custom assignment is used to support the user tag function.

Optionally, the configuration items of the interface node include configuration items configured to support interface information retrieval configuration.

One aspect of the present invention provides an apparatus for implementing multiple rounds of human-computer conversations, the apparatus comprising: a user configuration interface providing unit which provides a user configuration interface, wherein the user configuration interface providing unit provides a plurality of types of intention nodes in the user configuration interface; the intention unit construction unit is used for constructing an intention unit based on the operations of selecting an intention node, configuring the intention node and connecting the intention node on the user configuration interface by a first user; and the human-computer multi-turn conversation realization unit is used for realizing human-computer multi-turn conversation with the second user based on the intention unit so as to obtain the key intention of the second user in the human-computer multi-turn conversation and realize judgment and prediction of the conversation.

Optionally, the customized judgment value includes at least one of a task parameter value in the data received from the previous node, a parameter value in a parameter dictionary, and an interface return value, wherein the customized reference value includes at least one of a set value in the parameter dictionary, an interface return value, and a customized value, and wherein the customized logical relationship is a logical relationship between the customized judgment value selected by the first user according to the intention node and the customized reference value.

An aspect of the invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by one or more computing devices, causes the one or more computing devices to carry out any of the methods described above.

An aspect of the invention provides a multi-turn human-machine dialog system comprising one or more computing devices and one or more storage devices having a computer program recorded thereon, which, when executed by the one or more computing devices, causes the one or more computing devices to carry out any of the methods described above.

The invention can flexibly select the intention node, the configuration intention node and the connection intention node on the user configuration interface according to the self service requirement and the service experience of the first user, thereby constructing an intention unit for the dialogue robot program and facing the second user. Therefore, the conversation robot program can flexibly deal with the complex conversation situation of the user based on the intention unit facing the second user, effectively acquire the key intention of the user, and provide accurate information and service.

Drawings

The above and other objects and features of the present invention will become more apparent from the following description taken in conjunction with the accompanying drawings which illustrate, by way of example, an example in which:

FIG. 1 shows a flow diagram of a method of implementing a multi-turn conversation between humans and machines, according to an embodiment of the invention;

FIG. 2 illustrates a schematic diagram of building an intent unit in a user configuration interface, according to an embodiment of the present invention;

FIG. 3 shows a schematic diagram of a configuration of an input node according to an embodiment of the invention;

FIG. 4 shows a schematic diagram of a configuration of a decision node according to an embodiment of the invention;

FIG. 5 shows a schematic diagram of a configuration of an input node according to an embodiment of the invention;

FIG. 6 shows a schematic diagram of a configuration of an assignment node, according to an embodiment of the invention;

FIG. 7 shows a schematic diagram of a configuration of an interface node according to an embodiment of the invention;

FIG. 8 illustrates an apparatus for implementing multiple rounds of human-machine dialog, according to an embodiment of the present invention;

FIG. 9 shows an example of a multi-turn conversation of a human machine according to an embodiment of the invention;

FIG. 10A illustrates a build intent unit for a first round of dialog for the example of a human-machine multi-round dialog illustrated in FIG. 9;

FIG. 10B shows the configuration of the input nodes and the configuration of the judgment nodes of the construction intention unit shown in FIG. 10A;

fig. 10C shows the configuration of an output node corresponding to the parameter logic 1 and the configuration of an output node corresponding to the parameter logic 2 of the construction intention unit shown in fig. 10A and the configuration of the addition task parameter;

FIG. 10D illustrates a configuration interface for adding task parameters according to an embodiment of the invention;

FIG. 11A illustrates a further constructed intent unit for a second round of dialog for the example of a human-machine multi-round dialog illustrated in FIG. 9;

FIG. 11B shows the configuration of input nodes and the configuration of decision nodes of the construction intention unit shown in FIG. 11A;

fig. 11C shows a configuration of an output node corresponding to the parameter logic 1 and a configuration of an output node corresponding to the parameter logic 2 of the construction intention unit shown in fig. 11A;

FIG. 12A illustrates a further constructed intent unit for a third round of dialog for the example of a multi-round of human-machine dialog illustrated in FIG. 9;

FIG. 12B shows the configuration of the input nodes and the configuration of the judgment nodes of the construction intention unit shown in FIG. 12A;

fig. 12C shows a configuration of an output node corresponding to the parameter logic 1 and a configuration of an output node corresponding to the parameter logic 2 of the construction intention unit shown in fig. 12A;

FIG. 13A illustrates a further constructed intent unit for a fourth round of dialog for the example of a human-machine multi-round dialog illustrated in FIG. 9;

FIG. 13B shows the configuration of input nodes and the configuration of decision nodes of the construction intention unit shown in FIG. 13A;

fig. 13C shows a configuration of an output node corresponding to the parameter logic 1 and a configuration of an output node corresponding to the parameter logic 2 of the construction intention unit shown in fig. 13A;

FIG. 14A shows further constructed intent units for a fifth and sixth round of dialog for the example of a human-machine multi-round dialog shown in FIG. 9;

FIG. 14B illustrates a configuration of input nodes, assignment nodes, and decision nodes based on a fifth round of dialog with the build intent unit illustrated in FIG. 14A;

fig. 14C shows a configuration of an output node corresponding to the parameter logic 1, a configuration of an output node corresponding to the parameter logic 2 based on the fifth round of dialogue with the construction intention unit shown in fig. 14A;

FIG. 14D shows a configuration of input nodes and decision nodes based on a sixth round of dialog of the build intent unit shown in FIG. 14A;

fig. 14E shows the configuration of the output node corresponding to the parameter logic 1 and the configuration of the output node corresponding to the parameter logic 2 based on the sixth round of dialogue of the construction intention unit shown in fig. 14A;

FIG. 15A shows a further constructed intent unit for an eighth round of dialog for the example of a human-machine multi-round of dialog shown in FIG. 9;

FIG. 15B shows the configuration of the input nodes and the configuration of the judgment nodes of the construction intention unit shown in FIG. 15A;

fig. 15C shows a configuration of an output node corresponding to the parameter logic 1 and a configuration of an output node corresponding to the parameter logic 2 and an assignment node of the construction intention unit shown in fig. 15A;

FIG. 16A shows a further constructed intent unit for a ninth round of dialog for the example of a human-machine multi-round dialog shown in FIG. 9;

FIG. 16B shows the configuration of the input nodes and the configuration of the judgment nodes of the construction intention unit shown in FIG. 16A;

fig. 16C shows the configuration of an output node corresponding to the parameter logic 1 and the configuration of an output node corresponding to the parameter logic 2 and an assignment node of the construction intention unit shown in fig. 16A.

Detailed Description

The following description is provided with reference to the accompanying drawings to assist in a comprehensive understanding of exemplary embodiments of the invention as defined by the claims and their equivalents. The description includes various specific details to aid understanding, but these details are to be regarded as illustrative only. Thus, one of ordinary skill in the art will recognize that: various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Moreover, descriptions of well-known functions and constructions may be omitted for clarity and conciseness.

Fig. 1 shows a flow diagram of a method of implementing a multi-turn conversation between a human and a machine according to an embodiment of the invention.

Referring to fig. 1, in step S110, a user configuration interface is provided, wherein a plurality of types of intention nodes are provided in the user configuration interface.

Here, different types of intent nodes may be used to perform different operations. According to embodiments of the invention, the plurality of types of intent nodes include one or more of: the system comprises an input node, a judgment node, an output node, an assignment node and an interface node. These nodes will be described in more detail later.

In step S120, an intention unit is constructed based on operations of the first user selecting an intention node, configuring the intention node, and connecting the intention node on the user configuration interface.

That is, the first user may be the user who constructed the intent unit. For example, the first user may be a facilitator who provides information or services (e.g., a hotel order, a ticket order, an air order ticket, an information consultation, etc.).

Here, the first user may flexibly select to select the intention node, the configuration intention node and the connection intention node on the user configuration interface according to own business requirements and business experiences, thereby constructing a flexible intention unit for the dialogue robot program facing the second user.

For example, a first user may select, configure, and connect intent nodes on a user configuration interface according to a business context, such as historical data (e.g., data related to needs, habits, and/or preferences, etc.) of its target user (e.g., a second user using information or services provided by the first user), thereby building a flexible intent unit for the second user. However, the invention is not limited thereto, and the first user may select the intention node, configure the intention node, and connect the intention node on the user configuration interface according to any other method, thereby constructing a flexible intention unit for the second user. In the present invention, a single intent unit can perform a single intent man-machine multi-turn dialog.

In step S130, a human-computer multi-turn dialog with the second user is realized based on the intention unit to acquire a key intention of the second user in the human-computer multi-turn dialog, so as to realize judgment and prediction of the dialog.

Because the intention unit is flexibly configured by the first user as an intention unit facing the second user for the dialogue robot program, the dialogue robot program can flexibly and effectively realize man-machine multi-turn dialogue with the second user based on the intention unit so as to acquire the clear instruction of the second user.

As an example, a method of implementing a multi-turn conversation between humans may be performed by a conversation bot that provides services or information.

FIG. 2 shows a schematic diagram of building an intent unit in a user configuration interface, according to an embodiment of the invention.

Referring to fig. 2, in fig. 2, a user configuration interface 200 may include a first area 210 in which various types of intention nodes are arranged and a second area 220 for constructing an intention unit.

Here, note that the size and position of the first area 210 and the second area 220 may be arbitrarily set, and are not limited to the example in fig. 2. Further, although fig. 2 illustrates five types of intention nodes (i.e., input nodes, judgment nodes, output nodes, assignment nodes, and interface nodes), the intention nodes of the present invention are not limited to the above nodes or may include one or more of the input nodes, judgment nodes, output nodes, assignment nodes, and interface nodes. In addition, the intention nodes are all configured by the first user, so that personalized services or functions can be provided according to the needs of the first user.

Here, as an example, when the first user wants to construct an intent unit, the first user may select (e.g., drag, click, etc.) an intent node from the first region such that the selected corresponding intent node appears in the second region. In addition, the first user can flexibly configure the selected intention node, so that the intention node can execute respective functions or operations according to the flexible configuration, thereby acquiring the key intention of the second user and meeting the use requirement or intention requirement of the second user.

In addition, the first user may also connect the selected intention nodes, so that the connected intention nodes may perform operations corresponding to the intention unit in the logical relationship of the connection.

As a non-limiting example, when an intent node is selected, a Bezier curve may appear until the next intent node is dragged in, the pre-post nodes may automatically connect, and the nodes and arrows that the connection is complete may also change. However, the above examples are non-limiting, and the invention is not limited to the specific selection and connection of the intended nodes.

Since the first user can flexibly configure the intention unit for the second user according to a business scenario or actual needs, the dialogue robot program having the flexibly configured intention unit can effectively provide the second user with information or services in accordance with his intention.

Fig. 3 shows a schematic diagram of a configuration of an input node according to an embodiment of the invention.

Here, it is noted that the configuration of the input node shown in fig. 3 is illustrative, and the present invention is not limited thereto. The input nodes of the present invention may include one or more of the configuration items shown in fig. 3, or may include other configuration items.

In one embodiment of the invention, the configuration items of the input node include: the system comprises a triggering question for triggering a task and a question similar to the triggering question, a slot position and task parameters, wherein the slot position is set based on the triggering question and the question similar to the triggering question, and the task parameters are set based on the triggering question and/or the question similar to the triggering question.

Referring to fig. 3, configuration items of the input nodes may include a trigger question for triggering the task "hotel reservation" (e.g., the second user has a "please help me reserve a hotel" and "help me reserve a hotel") and a question similar to the trigger question (e.g., the second user has a "i want to reserve a hotel" and "help me reserve a hotel" dialog with the conversation bot).

Further, the configuration item of the input node may include a slot "hotel name" set based on a trigger question "please help me to book a hotel", and at this time, the information of the slot "hotel name" may be "a hotel".

In addition, the configuration items of the input nodes may further include task parameters (e.g., hotel name, check-in time, etc.) set based on the trigger query and/or a query similar to the trigger query.

In the process of configuring the input node by the first user, the first user can delete the configured image or add a new configuration item according to the situation, so that the configuration item can meet the use requirement of the second user.

Preferably, the configuration item of the input node may further include a jump question method for jumping to other input nodes. Therefore, even if the current dialog information of the second user is not matched with the current input node, the current dialog information of the second user can be jumped to the corresponding input node through a jump method for jumping to other input nodes, so that multiple rounds of man-machine dialogues can be normally carried out, and smooth use of the second user is guaranteed.

In one example, the initial input node does not exhibit a pre-jump question set.

Further, throughout the figures, the pen symbols may indicate that information may be entered and the trash can may indicate that it may be deleted.

Fig. 4 shows a schematic diagram of a configuration of a decision node according to an embodiment of the invention.

Here, it is noted that the configuration of the determination node shown in fig. 4 is illustrative, and the present invention is not limited thereto. The decision node of the present invention may include one or more of the configuration items shown in fig. 4, or may include other configuration items.

In one embodiment, the configuration item of the judgment node comprises parameter logic, wherein the parameter logic performs logic condition judgment on the value obtained by the last node and a known value or a custom value, and determines the next node of the judgment node based on the mutually exclusive judgment result.

Preferably, the parameter logic may include parameter logic corresponding to the slot being empty. Therefore, when the slot position of the second user is empty in the multi-round man-machine conversation, the conversation robot program can also provide a corresponding logic judgment mechanism, so that the multi-round man-machine conversation is smoothly carried out.

Referring to fig. 4, the configuration items of the decision node include parameter logic 1 and parameter logic 2. Here, the parameter logic 1 may correspond to a case where the "hotel name" is empty, and the parameter logic 2 may correspond to a case where the "hotel name" is the a hotel. The first user may save the configured parameter logic by pressing the button "save logic". At this time, the task of the determination node may be configured as "hotel". Here, the parameter logic 1 and the parameter logic 2 are parallel parameter logics.

For example, in a case where the second user has made a dialog with the dialog bot "please help me to book hotel a", the determination node may determine that the parameter logic 1 (i.e., hotel name ═ hotel a ]) is established, and then determine the next node according to the parameter logic 1. For another example, in a case where the second user has made a "please help me to book a hotel" conversation with the conversation robot program, the determination node may determine that the parameter logic 2 (i.e., the store name is null) is established, and then determine the next node according to the parameter logic 2.

In one example, the configuration items of the parameter logic may include a custom judgment value, a custom logic relationship, and a custom reference value, wherein the parameter logic is configured to determine whether the custom judgment value and the custom reference value satisfy the custom logic relationship, and determine a next node of the judgment node based on a result of the determination.

In other words, when the parameter logic determines that the custom judgment value and the custom reference value satisfy the custom logical relationship, then the next node of the judgment node is determined as the next node corresponding to the case that the custom judgment value and the custom reference value satisfy the custom logical relationship; when the parameter logic determines that the custom judgment value and the custom reference value do not satisfy the custom logic relationship, the next node of the judgment node is determined to be the next node corresponding to the condition that the custom judgment value and the custom reference value do not satisfy the custom logic relationship.

Here, the custom judgment value includes at least one of a task parameter value in data received from a previous node, a parameter value in a parameter dictionary, and an interface return value. Here, the parameter dictionary may be a pre-designed database including parameter values. Interface return values may be values or data sets returned from the interface node. Note that the parameter dictionary cannot use the logical relationship "equal" because it has a plurality of parameter values.

Further, the custom reference value may include at least one of a collective value, an interface return value, and a custom value in a parameter dictionary. The custom logical relationship is a logical relationship between a custom judgment value and a custom reference value set by the first user according to the intention node.

By way of non-limiting example, a custom logical relationship can include "empty," non-empty, "" equal to, "" not equal to, "" greater than, "" less than, "" belonging to, "" not belonging to, "" including, "" not including, "" and/or "or.

Fig. 5 shows a schematic diagram of a configuration of an input node according to an embodiment of the invention.

Here, it is noted that the configuration of the output node shown in fig. 5 is illustrative, and the present invention is not limited thereto. The output nodes of the present invention may include one or more of the configuration items shown in fig. 5, or may include other configuration items.

According to an embodiment of the present invention, the configuration item of the output node includes a configuration item that outputs the reply content to the second user in a predetermined form.

Referring to fig. 5, the output node is configured to perform the task "robot enquiry hotel name". Here, the first user may configure a plurality of pieces of reply content. For example, the first user may configure the reply content "ask you what hotel they want to subscribe? ". In one example, the first user may configure the reply content based on the second user's intent and the dialog context.

Further, referring to fig. 5, the predetermined form of outputting the reply content to the second user may include a picture, a voice, a link, a menu, and rich text.

In one example, when the second user performs multiple rounds of human-machine conversations with the conversation bot through the display screen of the bot to provide the information query service, the conversation bot may output the reply content of the conversation bot to the second user in the form of pictures, links, menus, and rich text through the display screen. In another example, when the second user performs multiple rounds of human-machine conversations with the conversation bot through the voice conversation module of the conversation bot to provide the customer service, the conversation bot may output the reply content of the conversation bot to the second user in the form of a voice through the voice conversation module.

However, the above examples are non-limiting, and any other predetermined form is also possible.

The text box of the content editor of the output node in the default state also supports adding the acquired task parameters and the output parameters of the interface node, namely the interface return value, and replying a more personalized word technique. For example, reply: good, { time }, { departure airport } flight } to { destination airport } has been reserved for you. (the contents in the symbol { } indicate the parameter values, in this example, { time }, { departure airport }, { destination airport } are task parameters, and { flight } is the interface return value.)

In other words, the output node may be configured to provide a multi-form reply based on the task parameters and the interface return values, thereby satisfying a wide variety of needs of the second user.

Fig. 6 shows a schematic diagram of a configuration of an assignment node according to an embodiment of the present invention.

Here, it is to be noted that the configuration of the assignment node shown in fig. 6 is illustrative, and the present invention is not limited thereto. The assignment nodes of the present invention may include one or more of the configuration items shown in fig. 6, or may include other configuration items.

According to an embodiment of the present invention, the configuration item of the assignment node may include at least one of a parameter reset and a custom assignment, wherein the parameter reset includes an alternation of slot information based on data received from a node immediately above the assignment node, and the custom assignment is used to support the user tag function. Because the user-defined assignment is configured to support the user tag function, the conversation robot program supports the service tag function at the scene node while supporting the service scene, so that the data deposition and feedback of the user portrait are realized.

Referring to FIG. 6, the assignment node is configured to perform the task "customer type". Here, the slot may be acquired by a previous node. In fig. 6, the parameter reset may include a hotel name or a check-in time.

Further, in FIG. 6, the custom assignment may be configured as a customer type for the second user, where the customer type is shown to include an intended user, a potential user, and an invalid customer, as examples. Thus, an efficient classification of the customer can be achieved in a plurality of rounds of the human-machine conversation of the second user.

Fig. 7 shows a schematic diagram of a configuration of an interface node according to an embodiment of the invention.

Here, it is noted that the configuration of the interface node shown in fig. 7 is illustrative, and the present invention is not limited thereto. The interface node of the present invention may include one or more of the configuration items shown in fig. 7, or may include other configuration items.

According to an embodiment of the present invention, the configuration items of the interface node include configuration items configured to support an interface information retrieval configuration.

Referring to fig. 7, the interface node may be configured to correspond to the task "query for empty room". Here, the configuration item "interface name" of the interface node may be configured to call an interface corresponding to the "interface name", and here, the configuration item "URL" of the interface node may be configured to call a uniform resource locator corresponding to the "URL". The configuration item "request type" of the interface node can be chosen to be of different types depending on the configuration of the first user, both HTTP _ GET/HTTP _ POST currently being provided.

Further, the first user may also set the input parameters and the output parameters according to a specification corresponding to the second user's intention to better provide the interface return values. For example, when the second user intends to book a hotel, the first user may configure through parameter values (e.g., information on hotel name, check-in time, check-in type, etc.) acquired from the user, and the configuration item on the output parameter may be configured based on the configuration item of the input parameter. (for example, calling the vacant room information corresponding to the satisfaction of the condition from the interface based on the information of the hotel name, the living time, the type of the living room, etc.)

Fig. 8 shows an apparatus for implementing multiple rounds of human-computer dialog, according to an embodiment of the invention.

Referring to fig. 8, an apparatus 800 for implementing a multi-turn human-machine dialog may include a user configuration interface providing unit 810, an intention unit constructing unit 820, and a multi-turn human-machine dialog implementing unit 830.

The user configuration interface providing unit 810 may provide a user configuration interface, wherein the user configuration interface providing unit 810 provides a plurality of types of intention nodes in the user configuration interface. The intention unit construction unit 820 may construct an intention unit based on operations of the first user selecting an intention node, configuring the intention node, and connecting the intention nodes on the user configuration interface. The human-machine multi-turn dialog implementation unit 830 may implement a human-machine multi-turn dialog with the second user based on the intention unit to obtain an explicit instruction of the second user.

In other words, the user configuration interface providing unit 810 may perform the step 110 described with reference to fig. 1, the intention unit constructing unit 820 may perform the step 120 described with reference to fig. 1, and the man-machine multi-turn dialog implementing unit 830 may perform the step 130 described with reference to fig. 1. Accordingly, the user configuration interfaces, various types of nodes described with reference to fig. 2 to 7 may also be applied to the apparatus 800 of fig. 8 for implementing multiple rounds of human-machine conversations. For the sake of brevity, the same description is not repeated here.

FIG. 9 shows an example of a multi-turn conversation of a human machine according to an embodiment of the invention.

Referring to fig. 9, an example of a human-machine multi-turn conversation (or a target conversation) is shown as an airline ticket reservation, however, the present invention is not limited thereto, and examples of the human-machine multi-turn conversation may also include a hotel reservation, a ticket reservation, an information consultation, and the like.

In fig. 9, the USER may represent a second USER who performs multiple rounds of a robot conversation with the conversation robot program ROBT.

An example of configuring the intention unit based on the example of the man-machine multi-turn dialog shown in fig. 9 will be described later with reference to fig. 10A. Note that fig. 9 to the illustrated examples are for illustrative purposes, and the configuration of the intention unit and the intention node of the present invention is not limited to the above examples. Further, it is to be understood that the configuration of the intention node described with reference to fig. 10A to 7 is an example of the intention node described with reference to fig. 3 to 7, and thus, the description of the configuration of the intention node described with reference to fig. 3 to 7 also applies to the configuration of the intention node described with reference to fig. 10A to 7. Therefore, for the brevity of description, repeated descriptions will be omitted.

FIG. 10A illustrates a build intent unit for a first round of dialog for the example of a human-machine multi-round dialog illustrated in FIG. 9; FIG. 10B shows the configuration of the input nodes and the configuration of the judgment nodes of the construction intention unit shown in FIG. 10A; fig. 10C shows the configuration of the output node corresponding to the parameter logic 1 and the configuration of the output node corresponding to the parameter logic 2 of the construction intention unit shown in fig. 10A and the configuration of the addition task parameter.

Fig. 9 shows an example of a man-machine multi-turn conversation in which the conversation content of the second user in the first turn of the conversation is "please help me to order an air ticket". The first user can construct the intention unit shown in fig. 10A based on the dialog content "please help me to order an air ticket" of the second user.

Referring to fig. 10B, the input node is configured to perform the task "ticket booking". The second user may configure the configuration item "trigger question" of the input node to include "please help me reserve an air ticket" and "help me reserve an air ticket", and configure the configuration item "similarity question" of the input node to include "order ticket". Here, the second user may also configure the slot "ticket number". The information of the slot may be "one" according to the conversation content "please help me to order one air ticket" of the second user. And under the condition that the conversation content of the second user is 'please help me to order an air ticket', the jump question method is not triggered. After configuring the input nodes, the first user may click on the determination to maintain the corresponding configuration.

In fig. 10B, the decision node may be configured to perform the task "ticket number decision". The second user may configure the parameter logic to include a parameter logic 1 and a parameter logic 2, wherein the parameter logic 1 is configured to indicate a case that the "ticket number is empty" and the parameter logic 2 is configured to indicate a case that the "ticket number belongs to the (parameter dictionary) digital dictionary". Here, for example, a numeric dictionary, which is an example of a parameter dictionary, may record information on the number of tickets.

Referring to fig. 10A and 10C, the intention unit may be configured to include two output nodes (i.e., an output node corresponding to parameter logic 1 and an output node corresponding to parameter logic 2).

In fig. 10C, the output node corresponding to parameter logic 1 may be configured to execute the task "parameter logic 1 ticket count". The output node corresponding to the parameter logic 1 may be configured to output "please confirm the number of tickets you need" to the second user. For example, the reply content may be output in one or more of the output forms shown in fig. 10C. Further, the output node corresponding to parameter logic 2 may be configured to execute the task "parameter logic 2 origin". The output node corresponding to parameter logic 2 may be configured to output "ask you for your place of departure is? ".

In connection with the example of fig. 9, when the dialog content of the second user is "please help me to order a flight ticket", the dialog bot may output to the first user the content "ask your place of departure is? ".

FIG. 10D illustrates a configuration interface to add task parameters according to an embodiment of the invention.

Referring to fig. 10D, an "add task parameter" may pop up when the second user clicks on the "task parameter". In "add task parameter", a parameter name may be configured, and a parameter dictionary (e.g., system number) may be configured from the built-in dictionary 1, the built-in dictionary 2, the user dictionary 1, and the user dictionary 2. In addition, a parameter dictionary can also be added through the parameter dictionary +.

FIG. 11A illustrates a further constructed intent unit for a second round of dialog for the example of a human-machine multi-round dialog illustrated in FIG. 9; FIG. 11B shows the configuration of input nodes and the configuration of decision nodes of the construction intention unit shown in FIG. 11A; fig. 11C shows the configuration of an output node corresponding to the parameter logic 1 and the configuration of an output node corresponding to the parameter logic 2 of the construction intention unit shown in fig. 11A.

The dialog content of the second user in the second round of the example of the man-machine multi-round dialog shown in fig. 9 is "beijing". The first user can construct the intention unit shown in fig. 11A based on the dialog content "beijing" of the second user. Specifically, the portion of the newly constructed intent unit of fig. 11A compared to fig. 10A may correspond to the portion in the block in fig. 11A.

Referring to fig. 11B, the input node is configured to perform a task of 'origin city'. The second user may configure the configuration item "trigger question" of the input node to include "beijing" and "help me to decide an air ticket" and configure the configuration item "similar question" of the input node to include "order ticket". Here, the second user may also configure a slot "origin," the information of which may be a system location (e.g., Beijing). In addition, the slot may further include a ticket number, and the information of the slot may be "one" according to the previous session content "please help me to order one air ticket" of the second user. And in the case that the conversation content of the second user is the content related to the air ticket, triggering the jump question method and jumping to the previous input node corresponding to the air ticket. After configuring the input nodes, the first user may click on the determination to maintain the corresponding configuration.

In fig. 11B, the determination node may be configured to perform a task "departure determination". The second user may configure the parameter logic to include a parameter logic 1 and a parameter logic 2, where the parameter logic 1 is configured to indicate a case where the "origin is empty", and the parameter logic 2 is configured to indicate a case where the "origin belongs to the (parameter dictionary) origin dictionary". Here, for example, the departure place dictionary may record information about the departure place.

Referring to fig. 11A and 11C, the intention unit may be configured to include two output nodes (i.e., an output node corresponding to parameter logic 1 and an output node corresponding to parameter logic 2).

In fig. 11C, the output node corresponding to the parameter logic 1 may be configured to perform the task "parameter logic 1 output". The output node corresponding to the parameter logic 1 may be configured to output "please confirm your departure place, tell me again" to the second user. For example, the reply content may be output in one or more of the output forms shown in fig. 11C. Further, the output node corresponding to parameter logic 2 may be configured to perform the task "parameter logic 2 output". An output node corresponding to parameter logic 2 may be configured to output to the second user "ask for which airport you depart from beijing? ".

In connection with the example of fig. 9, when the dialog content of the second round of dialog of the second user is "beijing", the dialog bot may output to the first user the content "ask for which airport you go from beijing? ".

FIG. 12A illustrates a further constructed intent unit for a third round of dialog for the example of a multi-round of human-machine dialog illustrated in FIG. 9; FIG. 12B shows the configuration of the input nodes and the configuration of the judgment nodes of the construction intention unit shown in FIG. 12A; fig. 12C shows the configuration of an output node corresponding to the parameter logic 1 and the configuration of an output node corresponding to the parameter logic 2 of the construction intention unit shown in fig. 12A.

The second user's conversation content in the third round of conversation of the example of the man-machine multi-round conversation shown in fig. 9 is "capital international airport". The first user may construct the intention unit shown in fig. 12A based on the dialog content "capital international airport" of the second user. Specifically, the portion of the newly constructed intent unit of fig. 12A compared to fig. 11A may correspond to the portion in the block in fig. 12A.

Referring to FIG. 12B, the input node is configured to perform the task "departure airport". The second user may configure the configuration item "trigger question" of the input node to include "capital international airport" and "i want to go to capital international airport" and configure the configuration item "similar question" of the input node to include "my departure airport is capital international airport". Here, the second user may also configure the slot "beijing airport". For example, the information of the slot may be an capital international airport according to the dialog content of the second user. In the case that the dialog content of the second user is the content related to the air ticket or the content related to the departure city, the jump inquiry is triggered and jumps to the previous input node corresponding to the air ticket or the input node related to the departure city. After configuring the input nodes, the first user may click on the determination to maintain the corresponding configuration.

In fig. 12B, the decision node may be configured to perform the task "departure airport decision". The second user may configure the parameter logic to include a parameter logic 1 and a parameter logic 2, wherein the parameter logic 1 is configured to indicate a case where the "origin is empty" or a case where the "beijing airport does not belong to the beijing airport dictionary". The parameter logic 2 is configured to indicate a case where "beijing airport belongs to the beijing airport dictionary". Here, for example, the beijing airport dictionary may record information about airports in beijing.

Referring to fig. 12A and 12C, the intention unit may be configured to include two output nodes (i.e., an output node corresponding to parameter logic 1 and an output node corresponding to parameter logic 2).

In fig. 12C, the output node corresponding to the parameter logic 1 may be configured to perform the task "parameter logic 1 output". The output node corresponding to parameter logic 1 may be configured to output "please confirm your departure airport, tell me again" to the second user. For example, the reply content may be output in one or more of the output forms shown in fig. 12C. Further, the output node corresponding to parameter logic 2 may be configured to perform the task "parameter logic 2 output". An output node corresponding to parameter logic 2 may be configured to output to the second user "ask for your destination is? ".

In connection with the example of fig. 9, when the dialog content of the third round of dialog of the second user is "capital international airport", the dialog bot may output to the first user the content "ask your destination to be? ".

FIG. 13A illustrates a further constructed intent unit for a fourth round of dialog for the example of a human-machine multi-round dialog illustrated in FIG. 9; FIG. 13B shows the configuration of input nodes and the configuration of decision nodes of the construction intention unit shown in FIG. 13A; fig. 13C shows the configuration of an output node corresponding to the parameter logic 1 and the configuration of an output node corresponding to the parameter logic 2 of the construction intention unit shown in fig. 13A.

Fig. 9 shows an example of a man-machine multi-turn conversation in which the second user's conversation content in the fourth turn of the conversation is "hang state". The first user may construct the intention unit shown in fig. 13A based on the dialog content "hang state" of the second user. Specifically, the portion of the newly constructed intent unit of fig. 13A compared to fig. 12A may correspond to the portion in the block in fig. 13A.

Referring to fig. 13B, the input node is configured to perform a task "destination city". The second user may configure the configuration item "trigger question" of the input node to include "hangzhou" and "i want to go to hangzhou" and configure the configuration item "similarity question" of the input node to include "my destination is hangzhou". Here, the second user may also configure a slot "destination". For example, the information of the slot may be Hangzhou according to the dialog content of the second user. In the case where the dialogue content of the second user is the content related to the air order ticket, the content related to the departure city, or the content related to the departure airport, the jump inquiry method is triggered and jumps to the previous input node corresponding to the air order ticket, the input node related to the departure city, or the input node related to the departure airport. After configuring the input nodes, the first user may click on the determination to maintain the corresponding configuration.

In fig. 13B, the determination node may be configured to perform a task "destination city determination". The second user may configure the parameter logic to include a parameter logic 1 and a parameter logic 2, wherein the parameter logic 1 is configured to indicate a case where the "destination is empty" or a case where the "destination does not belong to the destination dictionary". The parameter logic 2 is configured to indicate a case where "the destination belongs to the destination dictionary". Here, for example, the destination dictionary may record information about the destination.

Referring to fig. 13A and 13C, the intention unit may be configured to include two output nodes (i.e., an output node corresponding to parameter logic 1 and an output node corresponding to parameter logic 2).

In fig. 13C, the output node corresponding to the parameter logic 1 may be configured to perform the task "parameter logic 1 output". The output node corresponding to the parameter logic 1 may be configured to output "please confirm your destination, tell me again" to the second user. For example, the reply content may be output in one or more of the output forms shown in fig. 13C. Further, the output node corresponding to parameter logic 2 may be configured to perform the task "parameter logic 2 output". An output node corresponding to parameter logic 2 may be configured to output to the second user "ask which airport in hangzhou you go? ".

In connection with the example of fig. 9, when the dialog content of the fourth round of dialog of the second user is "hangzhou", the dialog bot may output to the first user the content "ask which airport of hangzhou you go? ".

FIG. 14A shows further constructed intent units for a fifth and sixth round of dialog for the example of a human-machine multi-round dialog shown in FIG. 9; FIG. 14B illustrates a configuration of input nodes, assignment nodes, and decision nodes based on a fifth round of dialog with the build intent unit illustrated in FIG. 14A; fig. 14C shows the configuration of the output node corresponding to the parameter logic 1, the configuration of the output node corresponding to the parameter logic 2 based on the fifth round of dialogue with the construction intention unit shown in fig. 14A.

Fig. 9 shows an example of a human-machine multi-turn conversation in which the conversation content of the second user in the fifth turn of conversation is "i'm misclassified, not to hang state, to go to shanghai". The first user can construct the intent unit shown in fig. 14A based on the dialog content of the second user "i'm misclassified, not to hang, go to shanghai". Specifically, since the intention of the second user is changed, the intention node is reset, and the process proceeds to the input node connected to the reset intention in fig. 14B.

Referring to fig. 14B, the input node is configured to perform a task "destination city". The second user may configure the configuration item "trigger question" of the input node to include "not go to hang, go to shanghai" and configure the configuration item "similarity question" of the input node to include "my destination is shanghai". Here, the second user may also configure a slot "destination". For example, the information of the slot may be shanghai according to the dialog content of the second user. In the case where the dialogue content of the second user is the content related to the air order ticket, the content related to the departure city, or the content related to the departure airport, the jump inquiry method is triggered and jumps to the previous input node corresponding to the air order ticket, the input node related to the departure city, or the input node related to the departure airport. After configuring the input nodes, the first user may click on the determination to maintain the corresponding configuration.

In fig. 14B, the assignment node located after the input node may be configured to perform the task "parameter reset". For example, the first user may reset the parameter to the destination based on the dialog content of the second user "i'm misclassified, go to Hangzhou, change to Shanghai". Further, according to the contents of the dialog of the second user, it can be determined by selecting which of the intended customer, the potential customer, and the invalid customer the second user belongs to in the "customer type". Since the classification function of the user can be realized by configuring the assignment node, more elaborate services can be provided for the user.

In fig. 14B, the determination node may be configured to perform a task "destination city determination". The second user may configure the parameter logic to include a parameter logic 1 and a parameter logic 2, wherein the parameter logic 1 is configured to indicate a case where the "destination is empty" or a case where the "destination does not belong to the destination dictionary". The parameter logic 2 is configured to indicate a case where "the destination belongs to the destination dictionary". Here, for example, the destination dictionary may record information about the destination.

Referring to fig. 14A and 14C, the intention unit may be configured to include two output nodes (i.e., an output node corresponding to a parameter logic 1 and an output node corresponding to a parameter logic 2).

In fig. 14C, the output node corresponding to the parameter logic 1 may be configured to perform the task "parameter logic 1 output". The output node corresponding to the parameter logic 1 may be configured to output "please confirm your destination, tell me again" to the second user. For example, the reply content may be output in one or more of the output forms shown in fig. 14C. Further, the output node corresponding to parameter logic 2 may be configured to perform the task "parameter logic 2 output". An output node corresponding to parameter logic 2 may be configured to output to the second user "ask for which airport in shanghai you go? ".

In connection with the example of fig. 9, when the dialog content of the fifth turn of dialog of the second user is "i'm misclassified, not going to hangzhou, change to shanghai", the dialog bot may output to the first user the content "ask for which airport in hangzhou you go? ".

Therefore, even if the intention of the second user is changed, the interactive robot program can ensure that the man-machine interaction proceeds flexibly and efficiently.

FIG. 14D shows a configuration of input nodes and decision nodes based on a sixth round of dialog of the build intent unit shown in FIG. 14A; fig. 14E shows the configuration of the output node corresponding to the parameter logic 1 and the configuration of the output node corresponding to the parameter logic 2 based on the sixth round of dialogue of the construction intention unit shown in fig. 14A.

The dialog content of the second user in the sixth round of dialog of the example of the man-machine multi-round dialog shown in fig. 9 is "shanghai sky river airport". The first user may further construct the intention unit shown in fig. 14A based on the dialog content "shanghai sky river airport" of the second user. Specifically, since the intention of the second user is not changed, the processing proceeds to an input node in parallel with the input node connected to the reset intention in fig. 14B.

Referring to fig. 14D, the input node is configured to execute a task "destination airport". The second user may configure the configuration item "trigger question" of the input node to include "i want to go to the shanghai river airport" and "shanghai river", and configure the configuration item "similarity question" of the input node to include "shanghai river airport". Here, the second user may also configure the slot "shanghai airport". For example, the information of the slot may be the shanghai sky river airport according to the dialog content of the second user. In the case where the dialogue content of the second user is the content related to the air order ticket, the content related to the departure city, the content related to the departure airport, or the content related to the destination, the jump inquiry is triggered and jumps to the previous input node corresponding to the air order ticket, the input node related to the departure city, the input node related to the departure airport, or the input node related to the destination. After configuring the input nodes, the first user may click on the determination to maintain the corresponding configuration.

In fig. 14D, the determination node subsequent to the input node may be configured to perform the task "destination airport determination". The second user may configure the parameter logic to include a parameter logic 1 and a parameter logic 2, wherein the parameter logic 1 is configured to indicate a case where the "shanghai airport is empty" or a case where the "shanghai airport does not belong to the shanghai airport dictionary". The parameter logic 2 is configured to indicate a case where "shanghai airport belongs to shanghai airport dictionary". Here, for example, the shanghai airport dictionary may record information about shanghai airports.

Referring to fig. 14A and 14E, the intention unit may be configured to include two output nodes (i.e., an output node corresponding to the parameter logic 1 and an output node corresponding to the parameter logic 2) following the determination node.

In fig. 14E, the output node corresponding to the parameter logic 1 may be configured to perform the task "parameter logic 1 output". The output node corresponding to the parameter logic 1 may be configured to output "sorry, no shanghai river found, please confirm your destination airport" to the second user. For example, the reply content may be output in one or more of the output forms shown in fig. 14E. Further, the output node corresponding to parameter logic 2 may be configured to perform the task "parameter logic 2 output". An output node corresponding to parameter logic 2 may be configured to output "ask for your departure time is? ".

In connection with the example of fig. 9, when the dialogue content of the sixth round of dialogue of the second user is "shanghai sky river airport", the dialogue robot program may output the content "sorry, not find shanghai sky river, please confirm your destination airport" to the second user based on the output node of the intention unit configured in fig. 14A, 14D, and 14E corresponding to the parameter logic 1.

Further, in connection with the example of fig. 9, when the dialog content of the seventh turn of dialog of the second user is "shanghai rainbow bridge airport", the dialog bot may output to the second user the content "ask you for your departure time is? ".

FIG. 15A shows a further constructed intent unit for an eighth round of dialog for the example of a human-machine multi-round of dialog shown in FIG. 9; FIG. 15B shows the configuration of the input nodes and the configuration of the judgment nodes of the construction intention unit shown in FIG. 15A; fig. 15C shows the configuration of the output node corresponding to the parameter logic 1 and the configuration of the output node corresponding to the parameter logic 2 and the assignment node of the construction intention unit shown in fig. 15A.

The dialog content of the second user in the eighth round of dialog of the example of the man-machine multi-round dialog shown in fig. 9 is "10 am tomorrow. The first user can construct the intention unit shown in fig. 15A based on the dialog content "10 am tomorrow" of the second user. Specifically, the portion of the newly constructed intent unit of fig. 15A compared to fig. 14A may correspond to the portion in the block in fig. 15A.

Referring to fig. 15B, the input node is configured to execute a task "departure time". The second user may configure the configuration item "trigger question" of the input node to include "10 am departure bar" tomorrow. Here, the second user may also configure the slot "departure time". For example, the information of the slot may be 10 am tomorrow according to the dialog content of the second user. In the case where the dialogue content of the second user is the content related to the air order ticket, the content related to the departure city, the content related to the departure airport, or the content related to the destination, the jump inquiry is triggered and jumps to the previous input node corresponding to the air order ticket, the input node related to the departure city, the input node related to the departure airport, or the input node related to the destination. After configuring the input nodes, the first user may click on the determination to maintain the corresponding configuration.

In fig. 15B, the determination node may be configured to perform the task "departure time determination". The second user may configure the parameter logic to include a parameter logic 1 and a parameter logic 2, where the parameter logic 1 is configured to indicate a case where the "departure time is empty" or a case where the "departure time does not belong to the time dictionary". The parameter logic 2 is configured to indicate a case where "departure time belongs to the time dictionary". Here, for example, the time dictionary may record information about time.

Referring to fig. 15A and 15C, the intention unit may be configured to include two output nodes (i.e., an output node corresponding to a parameter logic 1 and an output node corresponding to a parameter logic 2) and an interface node.

In fig. 15C, the output node corresponding to the parameter logic 1 may be configured to perform the task "parameter logic 1 output". The output node corresponding to the parameter logic 1 may be configured to output "sorry cannot determine your trigger time, please confirm your departure time" to the second user. For example, the reply content may be output in one or more of the output forms shown in fig. 15C. Further, the output node corresponding to parameter logic 2 may be configured to perform the task "parameter logic 2 output". The output node corresponding to the parameter logic 2 may be configured to output to the second user "find a flight to the shanghai rainbow bridge airport at 10 am tomorrow, capital international airport," ask you which to select "and an interface return value" southern AH123456 "returned from the interface node. The interface node has been described in detail above with reference to fig. 7 and will not be described again here.

In connection with the example of fig. 9, when the dialog content of the eighth round of dialog of the second user is "10 am tomorrow", the dialog bot may output to the first user the content "find the next flight from tomorrow am 10 am, capital international airport to shanghai rainbow bridge airport, ask you which to select" and the interface return value "south AH 123456" returned from the interface node, based on the intention unit configured in fig. 15A to 15C.

FIG. 16A shows a further constructed intent unit for a ninth round of dialog for the example of a human-machine multi-round dialog shown in FIG. 9; FIG. 16B shows the configuration of the input nodes and the configuration of the judgment nodes of the construction intention unit shown in FIG. 16A; fig. 16C shows the configuration of an output node corresponding to the parameter logic 1 and the configuration of an output node corresponding to the parameter logic 2 and an assignment node of the construction intention unit shown in fig. 16A.

The dialog content of the second user in the ninth round of dialog of the example of the man-machine multi-round dialog shown in fig. 9 is "southern AH123456 bar". The first user may construct the intent unit shown in fig. 16A based on the dialog content "south navigation AH123456 bar" of the second user. Specifically, the portion of the newly constructed intent unit of fig. 16A compared to fig. 15A may correspond to the portion in the block in fig. 16A.

Referring to FIG. 16B, the input node is configured to perform the task "flight". The second user may configure the configuration item "trigger question" of the input node to include "southern AH 123456". Here, the second user may also configure the slot "flight". For example, the information of the slot may be south navigation AH123456 according to the dialog content of the second user. In the case where the dialogue content of the second user is the content related to the air order ticket, the content related to the departure city, the content related to the departure airport, or the content related to the destination, the jump inquiry is triggered and jumps to the previous input node corresponding to the air order ticket, the input node related to the departure city, the input node related to the departure airport, or the input node related to the destination. After configuring the input nodes, the first user may click on the determination to maintain the corresponding configuration.

In fig. 16B, the decision node may be configured to perform the task "flight decision". The second user may configure the parameter logic to include a parameter logic 1 and a parameter logic 2, wherein the parameter logic 1 is configured to indicate a "flight empty" condition or a "flight not belonging to the interface return value" condition. The parameter logic 2 is configured to indicate a case where "flight belongs to the interface return value". Here, for example, the interface return value may record information about the flight.

Referring to fig. 16A and 16C, the intention unit may be configured to include two output nodes (i.e., an output node corresponding to a parameter logic 1 and an output node corresponding to a parameter logic 2) and an assignment node.

In fig. 16C, the output node corresponding to the parameter logic 1 may be configured to perform the task "parameter logic 1 output". The output node corresponding to parameter logic 1 may be configured to output "sorry selected among the optional flights, reconfirming your flight" to the second user. For example, the reply content may be output in one or more of the output forms shown in fig. 16C. Further, the output node corresponding to parameter logic 2 may be configured to perform the task "parameter logic 2 output". The output node corresponding to parameter logic 2 may be configured to output "good, south navigation AH123456 destined for your 20 th month 14 th 2020, 10 am 32 min, capital international airport to shanghai rainbow bridge airport" to the second user.

In connection with the example of fig. 9, when the dialog content of the ninth round of dialog of the second user is "south navigation AH 123456", the dialog bot may output to the first user the content "good, south navigation AH 123456" that is scheduled for you for 7/14/2020, 10 am 32 min, junior international airport to shanghai rainbow bridge airport "based on the intention unit configured in fig. 16A to 16C.

In addition, the first user may configure the assignment node to perform the task "airline intent tags". The assignment node may include parameter resets and customer flags. The second user may reconfigure the parameters to the destination and configure the user's customer flag to the intended user in the event that the second customer selects southward AH 123456.

Fig. 9 to 16C are only examples shown for illustrative purposes, and the contents of the man-machine conversation of the present invention are not limited to the specific contents in the above examples. Referring to fig. 9 to 16C, the first user may flexibly select, configure and connect intention nodes on the user configuration interface according to own business requirements and business experiences, thereby constructing an intention unit for the dialogue robot facing the second user.

The method and apparatus for implementing a multi-turn dialog between a human and a machine according to an exemplary embodiment of the present invention have been described above with reference to fig. 1 to 16C. However, it should be understood that: the methods used in fig. 1-7 and 9-16C may be implemented by software, hardware, firmware, or any combination thereof that performs the specified function, and the methods, apparatuses, systems, units, etc. used in fig. 8 may be configured as software, hardware, firmware, or any combination thereof, respectively, that performs the specified function. For example, these systems, devices, units, etc. may correspond to dedicated integrated circuits, to pure software code, or to a combination of software and hardware. Further, one or more functions implemented by these systems, apparatuses, or units, etc. may also be uniformly executed by components in a physical entity device (e.g., processor, client, server, etc.).

Further, the above-described method may be implemented by a computer program recorded on a computer-readable storage medium. For example, according to an exemplary embodiment of the present invention, a computer-readable storage medium may be provided, having stored thereon a computer program which, when executed by one or more computing devices, causes the one or more computing devices to implement any of the methods disclosed in the present application.

For example, the computer program, when executed by one or more computing devices, causes the one or more computing devices to perform the steps of: providing a user configuration interface, wherein a plurality of types of intention nodes are provided in the user configuration interface; constructing an intention unit based on operations of a first user for selecting an intention node, configuring the intention node and connecting the intention node on the user configuration interface; and realizing man-machine multi-turn conversation with the second user based on the intention unit so as to obtain the key intention of the second user in the man-machine multi-turn conversation and realize judgment and prediction of the conversation.

The computer program in the computer-readable storage medium may be executed in an environment deployed in a computer device such as a client, a host, a proxy apparatus, a server, etc., and it should be noted that the computer program may be further used to perform additional steps other than the above steps or perform more specific processing when the above steps are executed, and the content of the additional steps and the further processing is mentioned in the description of the related method and apparatus with reference to fig. 1 to 8, and therefore will not be described again here to avoid repetition.

It should be noted that the method and apparatus for implementing multiple rounds of human-computer conversations according to the exemplary embodiments of the present invention may be completely dependent on the execution of a computer program to implement the corresponding functions, wherein each unit of the apparatus or system corresponds to each step in the functional architecture of the computer program, so that the whole apparatus or system is called by a special software package (e.g., lib library) to implement the corresponding functions.

For example, a multi-turn human-machine dialog system is provided according to an embodiment of the invention comprising one or more computing devices and one or more storage devices, wherein the one or more storage devices have stored therein a computer program that, when executed by the one or more computing devices, causes the one or more computing devices to implement any of the methods disclosed herein. For example, causing the one or more computing devices to perform the steps of: providing a user configuration interface, wherein a plurality of types of intention nodes are provided in the user configuration interface; constructing an intention unit based on operations of a first user for selecting an intention node, configuring the intention node and connecting the intention node on the user configuration interface; and realizing man-machine multi-turn conversation with the second user based on the intention unit so as to obtain the key intention of the second user in the man-machine multi-turn conversation and realize judgment and prediction of the conversation.

In particular, the computing devices described above may be deployed in servers as well as on node devices in a distributed network environment. Further, the computing device apparatus may also include a video display (such as a liquid crystal display) and a user interaction interface (such as a keyboard, mouse, touch input device, etc.). All components of the computing device apparatus may be connected to each other via a bus and/or network.

The computing device here need not be a single device, but may be any collection of devices or circuits that can execute the instructions (or sets of instructions) described above, either individually or in combination. The computing device may also be part of an integrated control computing device or computing device manager, or may be configured as a portable electronic device that interfaces with local or remote (e.g., via wireless transmission).

The computing device for performing the training method or the named entity recognition method of the neural network according to the exemplary embodiments of the present invention may be a processor, and such a processor may include a Central Processing Unit (CPU), a Graphic Processing Unit (GPU), a programmable logic device, a dedicated processor, a microcontroller, or a microprocessor. By way of example, and not limitation, the processor may also include analog processors, digital processors, microprocessors, multi-core processors, processor arrays, network processors, and the like. The processor may execute instructions or code stored in one of the storage devices, which may also store data. Instructions and data may also be transmitted and received over a network via a network interface device, which may employ any known transmission protocol.

The storage device may be integral to the processor, e.g., having RAM or flash memory disposed within an integrated circuit microprocessor or the like. Further, the storage device may comprise a stand-alone device, such as an external disk drive, storage array, or other storage device usable by any database computing device. The storage device and the processor may be operatively coupled or may communicate with each other, such as through an I/O port, a network connection, etc., so that the processor can read files stored in the storage device.

It should be noted that the exemplary implementation of the present invention focuses on solving the problem that the current complicated dialog situation facing the user, the dialog robot program is often difficult to flexibly deal with, and the problem corresponding to the intention of the user cannot be effectively solved. In particular, the invention can flexibly select the intention node, the configuration intention node and the connection intention node on the user configuration interface according to the self business requirement and the business experience of the first user, thereby constructing an intention unit facing the second user for the dialogue robot program. Therefore, the conversation robot program can flexibly deal with the complex conversation situation of the user based on the intention unit facing the second user, and effectively acquire the final instruction of the user, thereby providing accurate information and service.

While exemplary embodiments of the present application have been described above, it should be understood that the above description is exemplary only, and not exhaustive, and that the present application is not limited to the exemplary embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the present application. Therefore, the protection scope of the present application shall be subject to the scope of the claims.

Claims

1. A method of implementing a multi-turn conversation between humans, the method comprising:

providing a user configuration interface, wherein a plurality of types of intention nodes are provided in the user configuration interface;

constructing an intention unit based on operations of a first user for selecting an intention node, configuring the intention node and connecting the intention node on the user configuration interface;

and realizing man-machine multi-turn conversation with the second user based on the intention unit so as to obtain the key intention of the second user in the man-machine multi-turn conversation and realize judgment and prediction of the conversation.

2. The method of claim 1, wherein the plurality of types of intent nodes comprise one or more of: the system comprises an input node, a judgment node, an output node, an assignment node and an interface node.

3. The method of claim 2, wherein the configuration items of the input nodes comprise:

a trigger question for triggering the task and a question similar to the trigger question;

a slot position, wherein the slot position is set based on a trigger question method and a question method similar to the trigger question method;

and the task parameters are set based on the triggering question and/or the similar question to the triggering question.

4. The method of claim 3, wherein the configuration items of the input nodes further comprise jump interrogatories for jumping to other input nodes.

5. The method according to claim 2, wherein the configuration item of the judgment node comprises parameter logic, wherein the parameter logic performs logical condition judgment on the value obtained by the previous node and the known value or the custom value, and determines the next node of the judgment node based on mutually exclusive judgment results.

6. The method of claim 5, wherein the configuration items of the parameter logic include custom judgment values, custom logic relationships, and custom reference values,

wherein the parameter logic is configured to determine whether the custom decision value and the custom reference value satisfy a custom logic relationship, and determine a next node of the decision node based on a result of the determination.

7. The method of claim 6, wherein the custom predicate value comprises at least one of a task parameter value in data received from a previous node, a parameter value in a parameter dictionary, and an interface return value,

wherein the custom reference value comprises at least one of a collective value, an interface return value, and a custom value in a parameter dictionary,

the user-defined logical relationship is a logical relationship between a user-defined judgment value selected by the first user according to the intention node and a user-defined reference value.

8. An apparatus for enabling multiple rounds of human-machine dialog, the apparatus comprising:

a user configuration interface providing unit which provides a user configuration interface, wherein the user configuration interface providing unit provides a plurality of types of intention nodes in the user configuration interface;

the intention unit construction unit is used for constructing an intention unit based on the operations of selecting an intention node, configuring the intention node and connecting the intention node on the user configuration interface by a first user;

and the human-computer multi-turn conversation realization unit is used for realizing human-computer multi-turn conversation with the second user based on the intention unit so as to obtain the key intention of the second user in the human-computer multi-turn conversation and realize judgment and prediction of the conversation.

9. A computer-readable storage medium having stored thereon a computer program that, when executed by one or more computing devices, causes the one or more computing devices to implement the method of any of claims 1-7.

10. A multi-turn human-machine dialog system comprising one or more computing devices and one or more storage devices having a computer program recorded thereon, which, when executed by the one or more computing devices, causes the one or more computing devices to carry out the method of any one of claims 1-7.